Feb 24, 2012 (CIDRAP News) – Two leading voices on the potential threat of lab-modified H5N1 viruses laid out their arguments about the human H5N1 fatality rate and undetected cases today and yesterday, with one group claiming "millions" likely have been infected and the other group saying current World Health Organization (WHO) fatality-rate estimates are about right.
One of the main flashpoints at a live discussion on Feb 3 hosted by the New York Academy of Sciences (NYAS) was a disagreement over how common undetected H5N1 infections are, a key consideration in determining the disease's case-fatality ratio (CFR). If many cases have gone unrecognized, the CFR is lower than the apparent 59%.
Two scientists who clashed over the issue expanded on their arguments: Peter Palese, PhD, a virologist at Mount Sinai School of Medicine in New York City, and colleagues published its analysis in Science yesterday. And today Michael T. Osterholm, PhD, MPH, director of the University of Minnesota's Center for Infectious Disease Research and Policy (CIDRAP) and a colleague published an analysis in mBio, the online journal of the American Society for Microbiology (ASM).
The two groups came to different conclusions about seroprevalence estimates, which measure antibodies to H5N1 in blood and are the main tool used to detect asymptomatic or subclinical cases.
Palese's team published somewhat higher projections based on separate analyses, including one that factored in studies following Hong Kong's 1997 outbreak, which involved a different strain of H5N1 virus than those seen in the past 8 years. Their overall seroprevalence rate was from 1% to 2%, whereas Osterholm and coauthor Nicholas Kelley, PhD, a research associate at CIDRAP, found a rate of 0.47%. (CIDRAP is the publisher of CIDRAP News.)
When both groups focused on studies that used WHO criteria, however, they came to a similar conclusion of about a 0.5% seropositivity rate.
Palese's group said that, given their findings, there are probably millions of undetected infections, a claim that H5N1 experts contacted by CIDRAP News rejected, saying it goes beyond what the scientific evidence suggests.
All the experts favored a more conservative take on H5N1 seroprevalence rates, with one specifically cautioning against optimism when dealing with H5N1 death rates.
Vagaries of seroprevalence studies
Though research groups have conducted a variety of seroprevalence studies to explore the levels of asymptomatic or mild H5N1 cases, interpreting the findings and comparing results is fraught with uncertainties, such as how long H5N1 antibodies persist and what antibody levels should be used to define a positive case.
Most seroprevalence studies since 2003 in potentially exposed populations, such as poultry workers, healthcare workers, and household and social contacts of confirmed cases, have found few people with antibody levels that suggest unrecognized infections. Some studies found no evidence of infection, and others found extremely low levels, with ranges that were typically less than 3%.
However, researchers have puzzled over higher levels seen in seroprevalence studies that were conducted in Hong Kong shortly after the first human H5N1 outbreak in 1997. Antibody levels ranged from 3% in government workers who responded to outbreaks to 38% in healthcare workers who were exposed to the H5N1 virus. Some experts have suggested the differences in the two patterns might be explained by genetic differences between the 1997 strain and the more recent viruses.
Palese's analysis
At the NYAS discussion and in other outlets, Palese argued that the WHO CFR of 59% is far too high, because only severe infections in hospitalized patients are counted. He cited 10 studies with at least 500 subjects in which subclinical infection rates ranged from 0.2% to 5.6%.
In their analysis for Science today, Palese and his colleagues included 20 seroprevalence studies, which had a total of 12,677 participants. They said their overall analysis suggests that 1% to 2% had evidence of earlier H5N1 infection.
When they grouped all the studies together that used WHO criteria for confirmed cases they found a seropositivity rate of 1.2% (95% confidence interval [CI], 0.6% to 2.1%). A separate analysis of studies that used the researchers' own criteria put the seropositivity rate at 1.9% (95% CI, 0.5% to 3.4%).
A sub-analysis of poultry workers using WHO criteria found a 1.4% seropositive rate overall but a 3.2% rate surrounding the 1997 Hong Kong outbreaks and a 0.5% rate when the 1997 studies were excluded.
The group used a random-effects approach to account for variation between studies. Palese's coauthors are Taia Wang, PhD, associate faculty member in Mount Sinai's department of microbiology, and Michael Parides, PhD, a biostatistician at Mount Sinai.
Taken together, the studies show a subclinical or mild infection rate that isn't accounted for in WHO's H5N1 fatality rate, the group wrote. "Thus, the true fatality rate for H5N1 influenza viruses is likely to be less than the frequently reported rate of more than 50%," according to the report.
Though it's not possible to determine an accurate H5N1 fatality rate based on the studies, the authors estimated that a 1% to 2% infection rate would yield millions of infected people. Palese's group also acknowledged that the number of H5N1 deaths could be underestimated.
They called for further, large-scale, standardized studies to get a better handle on the number of H5N1 infections. "This information is critical for calculation of a real fatality rate that is not solely based on hospitalized patients," the group wrote.
Osterholm's methodology
Meanwhile, Osterholm, who is a member of the federal advisory group that considered the bioterror implications of the lab-modified H5N1 transmission studies, argued at the NYAS discussion that Palese was basing his assessment on a selected number of H5N1 seroprevalence studies, and he contended that 13 studies using the WHO criteria found a rate of 0.469%.
The advisory group, the National Science Advisory Board for Biosecurity (NSABB) has recommended that key details of the studies be withheld from publication.
During the debate, Osterholm downplayed the seroprevalence issue, saying that even if the H5N1 virus was 20 times less virulent than it is now, it would still be more lethal than the 1918 flu virus.
In the analysis for mBio, Osterholm and Kelley identified 24 H5N1 seroprevalence studies, excluding three from the 1997 H5N1 outbreak, because more recent H5N1 viruses are genetically different. They noted that the 1997 infections aren't included in the WHO's global H5N1 case count and the 1997 strain is not recommended for inclusion in H5N1 vaccines.
Osterholm and Kelley's analysis focused on the 13 studies that used WHO screening criteria and were conducted within 4 months of human cases or within 6 months of poultry outbreaks. In the five of these studies that reported the range of serologic titers detected, no participants had evidence of H5N1 infection based on WHO criteria, and only 13 participants had neutralization titers between 1:10 and 1:40.
One of the studies found that most subjects had detectable titers below 1:80, but the researchers who conducted that study reevaluated the dilutions and concluded that the titers did not reflect detectable H5N1 antibodies, according to Osterholm and Kelley.
They said targeted surveillance involving people who were exposed to the virus has not turned up any additional mild H5N1 infections, and though the systems can't identify 100% of cases, data so far suggest the number of missed subclinical infections is likely to be relatively small.
Like the Palese group, Osterholm and Kelley also said it's important to consider the possibility of missed fatal H5N1 infections.
They wrote that the serologic evidence supports the current WHO estimate of a 30% to 80% CFR and emphasized that the CFR was only one of a number of factors the NSABB considered when making its recommendation.
"Given the global population and the current dynamics of population movement around the world, an H5N1 pandemic, even with a relatively low case-fatality rate, would be a truly catastrophic even," Osterholm and Kelley wrote.
Their analysis also took issue with another argument used in the H5N1 study debate: that vaccines and antivirals would play a key role in mitigating an H5N1 pandemic. Experience with the countermeasures during the 2009 H1N1 pandemic showed that the vaccine was produced too late to have a major public health impact and that there were global disparities in the use and availability of antivirals.
Osterholm has been an advocate for better, more effective, and more quickly produced flu vaccines.
Future discussions on the current controversy should answer critical questions, such as how to safely conduct the studies in mammals, how to share critical methods and findings with those who need to know, and how to safeguard the viruses from lab escapes, the article says.
"The current controversy provides a valuable opportunity for scientists and public policy experts to work together in creating this roadmap for the future," it concludes.
Experts respond
Yi Guan, MD, PhD, a virologist at Hong Kong University, told CIDRAP News in an e-mail that he believes H5N1 seroprevalence rates are likely to be lower than what many studies have reported. He said microneutralization assays can generate a lot of false positives, based on his lab experience comparing the method with classical virus neutralization assay.
Conducting serological surveys using microneutralization assays without doing parallel tests to confirm the findings can produce H5N1 seroconversion rates amplified "many, many times or log," he wrote.
Guan said he personally has doubts about whether there are any subclinical H5N1 cases, based on his own experience reviewing unpublished data.
Maria Van Kerkhove, PhD, with the Medical Research Council Centre for Outbreak Analysis and Modelling in the Department of Infectious Disease Epidemiology at Imperial College London, told CIDRAP News in an e-mail that she understands why the Palese and Osterholm groups conducted the analyses, but she strongly cautioned against overinterpreting individual H5N1 seroprevalence studies or reading too much into pooled analyses of studies.
She said pooling H5N1 seroprevalence studies should be viewed with caution, because several factors between studies can vary, such as lab methods used and sera collection timing. Van Kerkhove also noted that scientists have a poor understanding of what a seropositive result means in terms of infection.
Van Kerkhove led a group that published a recent systematic review of H5N1 seroprevalence studies in Public Library of Science (PLoS) One. She said the group, which included scientists from a host of global health organizations including the WHO, had considered doing a pooled analysis, but decided against it, due to the heterogeneity and limited comparability of the studies.
Any analysis of H5N1 seroprevalence studies should account for the strain differences between 1997 and those that circulated after 2002, she said.
She added that Palese's group seems to be overinterpreting the data, with their assumption that a 1% to 2% infection rate in exposed group translates into millions of missed infections.
"While I agree that we are missing both infections and deaths, I do not believe that we are 'missing millions' of infections," Van Kerhkove said. "The data do not support this."
Marc Lipsitch, PhD, professor of epidemiology at the Harvard School of Public Health and director of the Center for Communicable Disease Dynamics, told CIDRAP News that in his view, studies in most high-risk populations show low seroprevalence, and the risk is likely to be even lower for the majority of people in countries where the H5N1 virus circulates in birds.
He said that, given the concerns about the lack of specificity of serologic measurements and the perils of comparing different studies, "I think that really all we can say about the prevalence of H5N1 highly pathogenic virus infections in humans is that it is quite rare, though difficult to say how rare it is."
Lipsitch added that Palese group's contention that millions of people have been infected with the H5N1 virus "goes far beyond the evidence."
Even if the CFR is much lower than 60%, as Palese's group contends, the virus still has the capacity to cause a highly damaging pandemic, he said, adding that the 2009 pandemic was fatal in roughly 1 in 10,000 infected patients, yet caused a major strain on the health system.
"Even if we are missing 100 milder infections for every case we detect of H5N1—and I don't think we are—we would still be talking about a risk of death 60 times or so higher than the most recent pandemic," Lipsitch said.
He cautioned against planning responses to public health threats on best-case scenarios. The possibility that the Palese group's view that the H5N1 threat is overstated is incorrect should move the world to contain the threat, he added.
"In the situation where experts disagree, it is only responsible to plan for the possibility that the optimists are wrong," Lipsitch said.
Osterholm told CIDRAP News, "The article speaks for itself. The discussion now needs to move on from whether H5N1 poses a serious threat as either a Mother Nature–made or a man-made virus—as it clearly does—and move toward how scientists can share their flu-related results with those who need to know and toward how we can prevent these viruses from accidentally escaping labs."
Palese did not respond to a CIDRAP News request for comments.
Wang TT, Parides MK, Palese P. Seroevidence for H5N1 influenza infections in humans: meta-analysis. Science 2012 (published online Feb 23) [Abstract]
Osterholm MT, Kelley NS. Mammalian-transmissible H5N1 influenza: facts and perspective. mBio 2012 Feb 24;3(2):e00045-12 [Full text]
See also:
Feb 22 ASM press release
Feb 9 CIDRAP News story "Undetected H5N1 cases seem few, but questions persist"
Feb 3 CIDRAP News story "Live debate airs major divisions in H5N1 research battles"
Jan 24, 2011, PLoS One study